tests/test_half_precision.py

+import pytest
+import platform
+
+import numpy as np
+import pystencils as ps
+
+
+@pytest.mark.parametrize('target', (ps.Target.CPU, ps.Target.GPU))
+def test_half_precison(target):
+    if target == ps.Target.CPU:
+        if not platform.machine() in ['arm64', 'aarch64']:
+            pytest.xfail("skipping half precision test on non arm platform")
+
+        if 'clang' not in ps.cpu.cpujit.get_compiler_config()['command']:
+            pytest.xfail("skipping half precision because clang compiler is not used")
+
+    if target == ps.Target.GPU:
+        pytest.importorskip("cupy")
+
+    dh = ps.create_data_handling(domain_size=(10, 10), default_target=target)
+
+    f1 = dh.add_array("f1", values_per_cell=1, dtype=np.float16)
+    dh.fill("f1", 1.0, ghost_layers=True)
+    f2 = dh.add_array("f2", values_per_cell=1, dtype=np.float16)
+    dh.fill("f2", 2.0, ghost_layers=True)
+
+    f3 = dh.add_array("f3", values_per_cell=1, dtype=np.float16)
+    dh.fill("f3", 0.0, ghost_layers=True)
+
+    up = ps.Assignment(f3.center, f1.center + 2.1 * f2.center)
+
+    config = ps.CreateKernelConfig(target=dh.default_target, default_number_float=np.float32)
+    ast = ps.create_kernel(up, config=config)
+
+    kernel = ast.compile()
+
+    dh.run_kernel(kernel)
+    dh.all_to_cpu()
+
+    assert np.all(dh.cpu_arrays[f3.name] == 5.2)
+    assert dh.cpu_arrays[f3.name].dtype == np.float16

tests/test_indexed_kernels.py

+import sympy as sp
+import numpy as np
+import pytest
+
+import pystencils as ps
+from pystencils import Assignment, Field, CreateKernelConfig, create_kernel, Target
+from pystencils.transformations import filtered_tree_iteration
+from pystencils.typing import BasicType, FieldPointerSymbol, PointerType, TypedSymbol
+
+
+@pytest.mark.parametrize('target', [ps.Target.CPU, ps.Target.GPU])
+def test_indexed_kernel(target):
+    if target == Target.GPU:
+        pytest.importorskip("cupy")
+        import cupy as cp
+
+    arr = np.zeros((3, 4))
+    dtype = np.dtype([('x', int), ('y', int), ('value', arr.dtype)])
+    index_arr = np.zeros((3,), dtype=dtype)
+    index_arr[0] = (0, 2, 3.0)
+    index_arr[1] = (1, 3, 42.0)
+    index_arr[2] = (2, 1, 5.0)
+
+    indexed_field = Field.create_from_numpy_array('index', index_arr)
+    normal_field = Field.create_from_numpy_array('f', arr)
+    update_rule = Assignment(normal_field[0, 0], indexed_field('value'))
+
+    config = CreateKernelConfig(target=target, index_fields=[indexed_field])
+    ast = create_kernel([update_rule], config=config)
+    kernel = ast.compile()
+
+    if target == Target.CPU:
+        kernel(f=arr, index=index_arr)
+    else:
+        gpu_arr = cp.asarray(arr)
+        gpu_index_arr = cp.ndarray(index_arr.shape, dtype=index_arr.dtype)
+        gpu_index_arr.set(index_arr)
+        kernel(f=gpu_arr, index=gpu_index_arr)
+        arr = gpu_arr.get()
+    for i in range(index_arr.shape[0]):
+        np.testing.assert_allclose(arr[index_arr[i]['x'], index_arr[i]['y']], index_arr[i]['value'], atol=1e-13)
+
+
+@pytest.mark.parametrize('index_size', ("fixed", "variable"))
+@pytest.mark.parametrize('array_size', ("3D", "2D", "10, 12", "13, 17, 19"))
+@pytest.mark.parametrize('target', (ps.Target.CPU, ps.Target.GPU))
+@pytest.mark.parametrize('dtype', ("float64", "float32"))
+def test_indexed_domain_kernel(index_size, array_size, target, dtype):
+    dtype = BasicType(dtype)
+
+    f = ps.fields(f'f(1): {dtype.numpy_dtype.name}[{array_size}]')
+    g = ps.fields(f'g(1): {dtype.numpy_dtype.name}[{array_size}]')
+
+    index = TypedSymbol("index", dtype=BasicType(np.int16))
+    if index_size == "variable":
+        index_src = TypedSymbol("_size_src", dtype=BasicType(np.int16))
+        index_dst = TypedSymbol("_size_dst", dtype=BasicType(np.int16))
+    else:
+        index_src = 16
+        index_dst = 16
+    pointer_type = PointerType(dtype, const=False, restrict=True, double_pointer=True)
+    const_pointer_type = PointerType(dtype, const=True, restrict=True, double_pointer=True)
+
+    src = sp.IndexedBase(TypedSymbol(f"_data_{f.name}", dtype=const_pointer_type), shape=index_src)
+    dst = sp.IndexedBase(TypedSymbol(f"_data_{g.name}", dtype=pointer_type), shape=index_dst)
+
+    update_rule = [ps.Assignment(FieldPointerSymbol("f", dtype, const=True), src[index + 1]),
+                   ps.Assignment(FieldPointerSymbol("g", dtype, const=False), dst[index + 1]),
+                   ps.Assignment(g.center, f.center)]
+
+    ast = ps.create_kernel(update_rule, target=target)
+
+    code = ps.get_code_str(ast)
+    assert f"const {dtype.c_name} * RESTRICT _data_f = (({dtype.c_name} * RESTRICT const)(_data_f[index + 1]));" in code
+    assert f"{dtype.c_name} * RESTRICT  _data_g = (({dtype.c_name} * RESTRICT )(_data_g[index + 1]));" in code
+
+    if target == Target.CPU:
+        assert code.count("for") == f.spatial_dimensions + 1
+

tests/test_json_serializer.py

+"""
+Test the pystencils-specific JSON encoder and serializer as used in the Database class.
+"""
+
+import numpy as np
+import tempfile
+
+from pystencils.config import CreateKernelConfig
+from pystencils import Target, Field
+from pystencils.runhelper.db import Database, PystencilsJsonSerializer
+
+
+def test_json_serializer():
+
+    dtype = np.float32
+
+    index_arr = np.zeros((3,), dtype=dtype)
+    indexed_field = Field.create_from_numpy_array('index', index_arr)
+
+    # create pystencils config
+    config = CreateKernelConfig(target=Target.CPU, function_name='dummy_config', data_type=dtype,
+                                index_fields=[indexed_field])
+
+    # create dummy database
+    temp_dir = tempfile.TemporaryDirectory()
+    db = Database(file=temp_dir.name, serializer_info=('pystencils_serializer', PystencilsJsonSerializer))
+
+    db.save(params={'config': config}, result={'test': 'dummy'})

tests/test_jupyter_extensions.ipynb

+%% Cell type:code id: tags:
+
+``` python
+from pystencils.session import *
+```
+
+%% Cell type:code id: tags:
+
+``` python
+dh = ps.create_data_handling(domain_size=(256, 256), periodicity=True)
+c_field = dh.add_array('c')
+dh.fill("c", 0.0, ghost_layers=True)
+```
+
+%% Cell type:code id: tags:
+
+``` python
+for x in range(129):
+    for y in range(258):
+        dh.cpu_arrays['c'][x, y] = 1.0
+```
+
+%% Cell type:code id: tags:
+
+``` python
+plt.scalar_field(dh.cpu_arrays["c"])
+```
+
+%% Output
+
+    <matplotlib.image.AxesImage at 0x117081c10>
+
+
+
+%% Cell type:code id: tags:
+
+``` python
+ur = ps.Assignment(c_field[0, 0], c_field[1, 0])
+config = ps.CreateKernelConfig(target=dh.default_target, cpu_openmp=False, skip_independence_check=True)
+ast = ps.create_kernel(ur, config=config)
+kernel = ast.compile()
+```
+
+%% Cell type:code id: tags:
+
+``` python
+c_sync = dh.synchronization_function_cpu(['c'])
+```
+
+%% Cell type:code id: tags:
+
+``` python
+def timeloop(steps=10):
+    for i in range(steps):
+        c_sync()
+        dh.run_kernel(kernel)
+    return dh.gather_array('c')
+```
+
+%% Cell type:code id: tags:
+
+``` python
+ps.jupyter.set_display_mode('video')
+```
+
+%% Cell type:code id: tags:
+
+``` python
+ani = ps.plot.scalar_field_animation(timeloop, rescale=True, frames=12)
+ps.jupyter.display_animation(ani)
+```
+
+%% Output
+
+    <IPython.core.display.HTML object>
+
+%% Cell type:code id: tags:
+
+``` python
+ps.jupyter.set_display_mode('image_update')
+```
+
+%% Cell type:code id: tags:
+
+``` python
+ani = ps.plot.scalar_field_animation(timeloop, rescale=True, frames=12)
+ps.jupyter.display_animation(ani)
+```
+
+%% Output
+
+
+
+%% Cell type:code id: tags:
+
+``` python
+def grid_update_function(image):
+    for i in range(40):
+        c_sync()
+        dh.run_kernel(kernel)
+    return dh.gather_array('c')
+```
+
+%% Cell type:code id: tags:
+
+``` python
+animation = ps.jupyter.make_imshow_animation(dh.cpu_arrays["c"], grid_update_function, frames=300)
+```
+
+%% Output
+
+
+
+%% Cell type:code id: tags:
+
+``` python
+ps.jupyter.set_display_mode("video")
+ps.jupyter.set_display_mode("window")
+ps.jupyter.set_display_mode("image_update")
+ps.jupyter.activate_ipython()
+```
+
+%% Output
+
+    ---------------------------------------------------------------------------
+    ImportError                               Traceback (most recent call last)
+Cell     In[14], line 2
+          1 ps.jupyter.set_display_mode("video")
+    ----> 2 ps.jupyter.set_display_mode("window")
+          3 ps.jupyter.set_display_mode("image_update")
+          4 ps.jupyter.activate_ipython()
+File     ~/pystencils/pystencils/src/pystencils/jupyter.py:115, in set_display_mode(mode)
+        113     display_animation_func = display_as_html_video
+        114 elif animation_display_mode == 'window':
+    --> 115     ipython.magic("matplotlib qt")
+        116     display_animation_func = display_in_extra_window
+        117 elif animation_display_mode == 'image_update':
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/IPython/core/interactiveshell.py:2539, in InteractiveShell.magic(self, arg_s)
+       2537 magic_name, _, magic_arg_s = arg_s.partition(' ')
+       2538 magic_name = magic_name.lstrip(prefilter.ESC_MAGIC)
+    -> 2539 return self.run_line_magic(magic_name, magic_arg_s, _stack_depth=2)
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/IPython/core/interactiveshell.py:2417, in InteractiveShell.run_line_magic(self, magic_name, line, _stack_depth)
+       2415     kwargs['local_ns'] = self.get_local_scope(stack_depth)
+       2416 with self.builtin_trap:
+    -> 2417     result = fn(*args, **kwargs)
+       2419 # The code below prevents the output from being displayed
+       2420 # when using magics with decodator @output_can_be_silenced
+       2421 # when the last Python token in the expression is a ';'.
+       2422 if getattr(fn, magic.MAGIC_OUTPUT_CAN_BE_SILENCED, False):
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/IPython/core/magics/pylab.py:99, in PylabMagics.matplotlib(self, line)
+         97     print("Available matplotlib backends: %s" % backends_list)
+         98 else:
+    ---> 99     gui, backend = self.shell.enable_matplotlib(args.gui.lower() if isinstance(args.gui, str) else args.gui)
+        100     self._show_matplotlib_backend(args.gui, backend)
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/IPython/core/interactiveshell.py:3603, in InteractiveShell.enable_matplotlib(self, gui)
+       3599         print('Warning: Cannot change to a different GUI toolkit: %s.'
+       3600                 ' Using %s instead.' % (gui, self.pylab_gui_select))
+       3601         gui, backend = pt.find_gui_and_backend(self.pylab_gui_select)
+    -> 3603 pt.activate_matplotlib(backend)
+       3604 configure_inline_support(self, backend)
+       3606 # Now we must activate the gui pylab wants to use, and fix %run to take
+       3607 # plot updates into account
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/IPython/core/pylabtools.py:360, in activate_matplotlib(backend)
+        355 # Due to circular imports, pyplot may be only partially initialised
+        356 # when this function runs.
+        357 # So avoid needing matplotlib attribute-lookup to access pyplot.
+        358 from matplotlib import pyplot as plt
+    --> 360 plt.switch_backend(backend)
+        362 plt.show._needmain = False
+        363 # We need to detect at runtime whether show() is called by the user.
+        364 # For this, we wrap it into a decorator which adds a 'called' flag.
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/matplotlib/pyplot.py:271, in switch_backend(newbackend)
+        268 # have to escape the switch on access logic
+        269 old_backend = dict.__getitem__(rcParams, 'backend')
+    --> 271 backend_mod = importlib.import_module(
+        272     cbook._backend_module_name(newbackend))
+        274 required_framework = _get_required_interactive_framework(backend_mod)
+        275 if required_framework is not None:
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/importlib/__init__.py:126, in import_module(name, package)
+        124             break
+        125         level += 1
+    --> 126 return _bootstrap._gcd_import(name[level:], package, level)
+File     <frozen importlib._bootstrap>:1204, in _gcd_import(name, package, level)
+File     <frozen importlib._bootstrap>:1176, in _find_and_load(name, import_)
+File     <frozen importlib._bootstrap>:1147, in _find_and_load_unlocked(name, import_)
+File     <frozen importlib._bootstrap>:690, in _load_unlocked(spec)
+File     <frozen importlib._bootstrap_external>:940, in exec_module(self, module)
+File     <frozen importlib._bootstrap>:241, in _call_with_frames_removed(f, *args, **kwds)
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/matplotlib/backends/backend_qt5agg.py:7
+          4 from .. import backends
+          6 backends._QT_FORCE_QT5_BINDING = True
+    ----> 7 from .backend_qtagg import (    # noqa: F401, E402 # pylint: disable=W0611
+          8     _BackendQTAgg, FigureCanvasQTAgg, FigureManagerQT, NavigationToolbar2QT,
+          9     FigureCanvasAgg, FigureCanvasQT)
+         12 @_BackendQTAgg.export
+         13 class _BackendQT5Agg(_BackendQTAgg):
+         14     pass
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/matplotlib/backends/backend_qtagg.py:9
+          5 import ctypes
+          7 from matplotlib.transforms import Bbox
+    ----> 9 from .qt_compat import QT_API, _enum
+         10 from .backend_agg import FigureCanvasAgg
+         11 from .backend_qt import QtCore, QtGui, _BackendQT, FigureCanvasQT
+File     /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/matplotlib/backends/qt_compat.py:135
+        133         break
+        134     else:
+    --> 135         raise ImportError(
+        136             "Failed to import any of the following Qt binding modules: {}"
+        137             .format(", ".join(_ETS.values())))
+        138 else:  # We should not get there.
+        139     raise AssertionError(f"Unexpected QT_API: {QT_API}")
+    ImportError: Failed to import any of the following Qt binding modules: PyQt6, PySide6, PyQt5, PySide2

tests/test_logarithm.py

+import pytest
+import numpy as np
+import sympy as sp
+
+import pystencils as ps
+
+
+@pytest.mark.parametrize('dtype', ["float64", "float32"])
+def test_log(dtype):
+    a = sp.Symbol("a")
+    x = ps.fields(f'x: {dtype}[1d]')
+
+    assignments = ps.AssignmentCollection({x.center(): sp.log(a)})
+
+    ast = ps.create_kernel(assignments)
+    code = ps.get_code_str(ast)
+    kernel = ast.compile()
+
+    # ps.show_code(ast)
+
+    if dtype == "float64":
+        assert "float" not in code
+
+    array = np.zeros((10,), dtype=dtype)
+    kernel(x=array, a=100)
+    assert np.allclose(array, 4.60517019)

pystencils_tests/test_loop_cutting.py → tests/test_loop_cutting.py

@@ -29,10 +29,6 @@ def offsets_in_plane(normal_plane, offset_int, dimension):
     return result
 
 
-# TODO this fails because the condition of the Conditional is not simplified anymore:
-# TODO: ---> transformation.simplify_conditionals
-# TODO this should be fixed
-@pytest.mark.xfail
 def test_staggered_iteration():
     dim = 2
     f_arr = np.arange(5**dim).reshape([5]*dim).astype(np.float64)

pystencils_tests/test_math_functions.py → tests/test_math_functions.py

@@ -2,6 +2,7 @@ import pytest
 import sympy as sp
 import numpy as np
 import pystencils as ps
+from pystencils.fast_approximation import fast_division
 
 
 @pytest.mark.parametrize('dtype', ["float64", "float32"])
@@ -9,7 +10,7 @@ import pystencils as ps
 @pytest.mark.parametrize('target', [ps.Target.CPU, ps.Target.GPU])
 def test_two_arguments(dtype, func, target):
     if target == ps.Target.GPU:
-        pytest.importorskip("pycuda")
+        pytest.importorskip("cupy")
     dh = ps.create_data_handling(domain_size=(10, 10), periodicity=True, default_target=target)
 
     x = dh.add_array('x', values_per_cell=1, dtype=dtype)
@@ -38,11 +39,11 @@ def test_two_arguments(dtype, func, target):
 
 
 @pytest.mark.parametrize('dtype', ["float64", "float32"])
-@pytest.mark.parametrize('func', [sp.sin, sp.cos, sp.sinh, sp.cosh, sp.atan])
+@pytest.mark.parametrize('func', [sp.sin, sp.cos, sp.sinh, sp.cosh, sp.atan, sp.floor, sp.ceiling])
 @pytest.mark.parametrize('target', [ps.Target.CPU, ps.Target.GPU])
 def test_single_arguments(dtype, func, target):
     if target == ps.Target.GPU:
-        pytest.importorskip("pycuda")
+        pytest.importorskip("cupy")
     dh = ps.create_data_handling(domain_size=(10, 10), periodicity=True, default_target=target)
 
     x = dh.add_array('x', values_per_cell=1, dtype=dtype)
@@ -57,7 +58,8 @@ def test_single_arguments(dtype, func, target):
     ast = ps.create_kernel(up, config=config)
     code = ps.get_code_str(ast)
     if dtype == 'float32':
-        assert func.__name__.lower() in code
+        func_name = func.__name__.lower() if func is not sp.ceiling else "ceil"
+        assert func_name in code
     kernel = ast.compile()
 
     dh.all_to_gpu()
@@ -66,3 +68,47 @@ def test_single_arguments(dtype, func, target):
 
     np.testing.assert_allclose(dh.gather_array("x")[0, 0], float(func(1.0).evalf()),
                                rtol=10**-3 if dtype == 'float32' else 10**-5)
+
+
+@pytest.mark.parametrize('a', [sp.Symbol('a'), ps.fields('a:  float64[2d]').center])
+def test_avoid_pow(a):
+    x = ps.fields('x:  float64[2d]')
+
+    up = ps.Assignment(x.center_vector[0], 2 * a ** 2 / 3)
+    ast = ps.create_kernel(up)
+
+    code = ps.get_code_str(ast)
+
+    assert "pow" not in code
+
+
+def test_avoid_pow_fast_div():
+    x = ps.fields('x:  float64[2d]')
+    a = ps.fields('a:  float64[2d]').center
+
+    up = ps.Assignment(x.center_vector[0], fast_division(1, (a**2)))
+    ast = ps.create_kernel(up, config=ps.CreateKernelConfig(target=ps.Target.GPU))
+    # ps.show_code(ast)
+
+    code = ps.get_code_str(ast)
+
+    assert "pow" not in code
+
+
+def test_avoid_pow_move_constants():
+    # At the end of the kernel creation the function move_constants_before_loop will be called
+    # This function additionally contains substitutions for symbols with the same value
+    # Thus it simplifies the equations again
+    x = ps.fields('x:  float64[2d]')
+    a, b, c = sp.symbols("a, b, c")
+
+    up = [ps.Assignment(a, 0.0),
+          ps.Assignment(b, 0.0),
+          ps.Assignment(c, 0.0),
+          ps.Assignment(x.center_vector[0], a**2/18 - a*b/6 - a/18 + b**2/18 + b/18 - c**2/36)]
+    ast = ps.create_kernel(up)
+
+    code = ps.get_code_str(ast)
+    ps.show_code(ast)
+
+    assert "pow" not in code

tests/test_modulo.py

+import pytest
+
+import numpy as np
+import sympy as sp
+import pystencils as ps
+from pystencils.astnodes import LoopOverCoordinate, Conditional, Block, SympyAssignment
+
+SLICE_LIST = [False,
+              ps.make_slice[1:-1:2, 1:-1:2],
+              ps.make_slice[2:-1:2, 4:-1:7],
+              ps.make_slice[4:-1:2, 5:-1:2],
+              ps.make_slice[3:-1:4, 7:-1:3]]
+
+
+@pytest.mark.parametrize('target', [ps.Target.CPU, ps.Target.GPU])
+@pytest.mark.parametrize('iteration_slice', SLICE_LIST)
+def test_mod(target, iteration_slice):
+    if target == ps.Target.GPU:
+        pytest.importorskip("cupy")
+    dh = ps.create_data_handling(domain_size=(51, 51), periodicity=True, default_target=target)
+
+    loop_ctrs = [LoopOverCoordinate.get_loop_counter_symbol(i) for i in range(dh.dim)]
+    cond = [sp.Eq(sp.Mod(loop_ctrs[i], 2), 1) for i in range(dh.dim)]
+
+    field = dh.add_array("a", values_per_cell=1)
+
+    eq_list = [SympyAssignment(field.center, 1.0)]
+
+    if iteration_slice:
+        config = ps.CreateKernelConfig(target=dh.default_target, iteration_slice=iteration_slice)
+        assign = eq_list
+    else:
+        assign = [Conditional(sp.And(*cond), Block(eq_list))]
+        config = ps.CreateKernelConfig(target=dh.default_target)
+
+    kernel = ps.create_kernel(assign, config=config).compile()
+
+    dh.fill(field.name, 0, ghost_layers=True)
+
+    if config.target == ps.enums.Target.GPU:
+        dh.to_gpu(field.name)
+
+    dh.run_kernel(kernel)
+
+    if config.target == ps.enums.Target.GPU:
+        dh.to_cpu(field.name)
+
+    result = dh.gather_array(field.name, ghost_layers=True)
+
+    assert np.all(result[iteration_slice] == 1.0)
+

pystencils_tests/test_move_constant_before_loop.py → tests/test_move_constant_before_loop.py

@@ -25,7 +25,40 @@ def test_symbol_renaming():
 
     loops = block.atoms(LoopOverCoordinate)
     assert len(loops) == 2
+    assert len(block.args[1].body.args) == 1
+    assert len(block.args[3].body.args) == 2
+
     for loop in loops:
-        assert len(loop.body.args) == 1
         assert len(loop.parent.args) == 4  # 2 loops + 2 subexpressions
-        assert loop.parent.args[0].lhs.name != loop.parent.args[1].lhs.name
+        assert loop.parent.args[0].lhs.name != loop.parent.args[2].lhs.name
+
+
+def test_keep_order_of_accesses():
+    f = ps.fields("f: [1D]")
+    x = TypedSymbol("x", np.float64)
+    n = 5
+
+    loop = LoopOverCoordinate(Block([SympyAssignment(x, f[0]),
+                                     SympyAssignment(f[1], 2 * x)]),
+                              0, 0, n)
+    block = Block([loop])
+
+    ps.transformations.resolve_field_accesses(block)
+    new_loops = ps.transformations.cut_loop(loop, [n - 1])
+    ps.transformations.move_constants_before_loop(new_loops.args[1])
+
+    kernel_func = ps.astnodes.KernelFunction(
+        block, ps.Target.CPU, ps.Backend.C, ps.cpu.cpujit.make_python_function, None
+    )
+    kernel = kernel_func.compile()
+
+    print(ps.show_code(kernel_func))
+
+    f_arr = np.ones(n + 1)
+    kernel(f=f_arr)
+
+    print(f_arr)
+
+    assert np.allclose(f_arr, np.array([
+        1, 2, 4, 8, 16, 32
+    ]))

tests/test_phasefield_dentritic_3D.ipynb

+%% Cell type:code id: tags:
+
+``` python
+import pytest
+pytest.importorskip('cupy')
+```
+
+%% Output
+
+    <module 'cupy' from '/home/markus/Python311/lib/python3.11/site-packages/cupy/__init__.py'>
+
+%% Cell type:code id: tags:
+
+``` python
+from pystencils.session import *
+sp.init_printing()
+frac = sp.Rational
+```
+
+%% Cell type:markdown id: tags:
+
+# Phase-field simulation of dentritic solidification in 3D
+
+This notebook tests the model presented in the dentritic growth tutorial in 3D.
+
+%% Cell type:code id: tags:
+
+``` python
+target = ps.Target.GPU
+gpu = target == ps.Target.GPU
+domain_size = (25, 25, 25) if 'is_test_run' in globals() else (300, 300, 300)
+
+dh = ps.create_data_handling(domain_size=domain_size, periodicity=True, default_target=target)
+φ_field = dh.add_array('phi', latex_name='φ')
+φ_field_tmp = dh.add_array('phi_tmp', latex_name='φ_tmp')
+φ_delta_field = dh.add_array('phidelta', latex_name='φ_D')
+t_field = dh.add_array('T')
+t_field_tmp = dh.add_array('T_tmp')
+```
+
+%% Cell type:code id: tags:
+
+``` python
+ε, m, δ, j, θzero, α, γ, Teq, κ, τ = sp.symbols("ε m δ j θ_0 α γ T_eq κ τ")
+εb = sp.Symbol("\\bar{\\epsilon}")
+discretize = ps.fd.Discretization2ndOrder(dx=0.03, dt=1e-5)
+
+φ = φ_field.center
+T = t_field.center
+d = ps.fd.Diff
+
+def f(φ, m):
+    return φ**4 / 4 - (frac(1, 2) - m/3) * φ**3 + (frac(1,4)-m/2)*φ**2
+
+
+
+bulk_free_energy_density = f(φ, m)
+interface_free_energy_density = ε ** 2 / 2 * (d(φ, 0) ** 2 + d(φ, 1) ** 2 + d(φ, 2) ** 2)
+```
+
+%% Cell type:markdown id: tags:
+
+Here comes the major change, that has to be made for the 3D model: $\epsilon$ depends on the interface normal, which can not be computed simply as atan() as in the 2D case
+
+%% Cell type:code id: tags:
+
+``` python
+n = sp.Matrix([d(φ, i) for i in range(3)])
+nLen = sp.sqrt(sum(n_i**2 for n_i in n))
+n = n / nLen
+nVal = sum(n_i**4 for n_i in n)
+σ = δ * nVal
+
+εVal = εb * (1 + σ)
+εVal
+```
+
+%% Output
+
+    $\displaystyle \bar{\epsilon} \left(δ \left(\frac{{\partial_{0} {φ}_{(0,0,0)}}^{4}}{\left({\partial_{0} {φ}_{(0,0,0)}}^{2} + {\partial_{1} {φ}_{(0,0,0)}}^{2} + {\partial_{2} {φ}_{(0,0,0)}}^{2}\right)^{2}} + \frac{{\partial_{1} {φ}_{(0,0,0)}}^{4}}{\left({\partial_{0} {φ}_{(0,0,0)}}^{2} + {\partial_{1} {φ}_{(0,0,0)}}^{2} + {\partial_{2} {φ}_{(0,0,0)}}^{2}\right)^{2}} + \frac{{\partial_{2} {φ}_{(0,0,0)}}^{4}}{\left({\partial_{0} {φ}_{(0,0,0)}}^{2} + {\partial_{1} {φ}_{(0,0,0)}}^{2} + {\partial_{2} {φ}_{(0,0,0)}}^{2}\right)^{2}}\right) + 1\right)$
+                   ⎛  ⎛                            4
+                   ⎜  ⎜                 D(φ[0,0,0])
+    \bar{\epsilon}⋅⎜δ⋅⎜───────────────────────────────────────────── + ───────────
+                   ⎜  ⎜                                            2
+                   ⎜  ⎜⎛           2              2              2⎞    ⎛
+                   ⎝  ⎝⎝D(φ[0,0,0])  + D(φ[0,0,0])  + D(φ[0,0,0]) ⎠    ⎝D(φ[0,0,0]
+    
+                     4                                               4
+          D(φ[0,0,0])                                     D(φ[0,0,0])
+    ────────────────────────────────── + ─────────────────────────────────────────
+                                     2
+     2              2              2⎞    ⎛           2              2
+    )  + D(φ[0,0,0])  + D(φ[0,0,0]) ⎠    ⎝D(φ[0,0,0])  + D(φ[0,0,0])  + D(φ[0,0,0]
+    
+        ⎞    ⎞
+        ⎟    ⎟
+    ────⎟ + 1⎟
+       2⎟    ⎟
+     2⎞ ⎟    ⎟
+    ) ⎠ ⎠    ⎠
+
+%% Cell type:code id: tags:
+
+``` python
+def m_func(temperature):
+    return (α / sp.pi) * sp.atan(γ * (Teq - temperature))
+```
+
+%% Cell type:code id: tags:
+
+``` python
+substitutions = {m: m_func(T),
+                 ε: εVal}
+
+fe_i = interface_free_energy_density.subs(substitutions)
+fe_b = bulk_free_energy_density.subs(substitutions)
+
+μ_if = ps.fd.expand_diff_full(ps.fd.functional_derivative(fe_i, φ), functions=[φ])
+μ_b = ps.fd.expand_diff_full(ps.fd.functional_derivative(fe_b, φ), functions=[φ])
+```
+
+%% Cell type:code id: tags:
+
+``` python
+dF_dφ = μ_b + sp.Piecewise((μ_if, nLen**2 > 1e-10), (0, True))
+```
+
+%% Cell type:code id: tags:
+
+``` python
+parameters = {
+    τ: 0.0003,
+    κ: 1.8,
+    εb: 0.01,
+    δ: 0.3,
+    γ: 10,
+    j: 6,
+    α: 0.9,
+    Teq: 1.0,
+    θzero: 0.2,
+    sp.pi: sp.pi.evalf()
+}
+parameters
+```
+
+%% Output
+
+
+    $\displaystyle \left\{ \pi : 3.14159265358979, \  T_{eq} : 1.0, \  \bar{\epsilon} : 0.01, \  j : 6, \  α : 0.9, \  γ : 10, \  δ : 0.3, \  θ_{0} : 0.2, \  κ : 1.8, \  τ : 0.0003\right\}$
+    {π: 3.14159265358979, T_eq: 1.0, \bar{\epsilon}: 0.01, j: 6, α: 0.9, γ: 10, δ:
+     0.3, θ₀: 0.2, κ: 1.8, τ: 0.0003}
+
+%% Cell type:code id: tags:
+
+``` python
+dφ_dt = - dF_dφ / τ
+assignments = [
+    ps.Assignment(φ_delta_field.center, discretize(dφ_dt.subs(parameters))),
+]
+φEqs = ps.simp.sympy_cse_on_assignment_list(assignments)
+φEqs.append(ps.Assignment(φ_field_tmp.center, discretize(ps.fd.transient(φ) - φ_delta_field.center)))
+
+
+temperatureEvolution = -ps.fd.transient(T) + ps.fd.diffusion(T, 1) + κ * φ_delta_field.center
+temperatureEqs = [
+    ps.Assignment(t_field_tmp.center, discretize(temperatureEvolution.subs(parameters)))
+]
+```
+
+%% Cell type:code id: tags:
+
+``` python
+φ_kernel = ps.create_kernel(φEqs, cpu_openmp=4, target=target).compile()
+temperatureKernel = ps.create_kernel(temperatureEqs, cpu_openmp=4, target=target).compile()
+```
+
+%% Cell type:code id: tags:
+
+``` python
+def time_loop(steps):
+    φ_sync = dh.synchronization_function(['phi'], target=target)
+    temperature_sync = dh.synchronization_function(['T'], target=target)
+    dh.all_to_gpu()
+    for t in range(steps):
+        φ_sync()
+        dh.run_kernel(φ_kernel)
+        temperature_sync()
+        dh.run_kernel(temperatureKernel)
+        dh.swap(φ_field.name, φ_field_tmp.name)
+        dh.swap(t_field.name, t_field_tmp.name)
+    dh.all_to_cpu()
+
+
+def init(nucleus_size=np.sqrt(5)):
+    for b in dh.iterate():
+        x, y, z = b.cell_index_arrays
+        x, y, z = x - b.shape[0] // 2, y - b.shape[1] // 2, z - b.shape[2] // 2
+        b['phi'].fill(0)
+        b['phi'][(x ** 2 + y ** 2 + z ** 2) < nucleus_size ** 2] = 1.0
+        b['T'].fill(0.0)
+
+
+def plot(slice_obj=ps.make_slice[:, :, 0.5]):
+    plt.subplot(1, 3, 1)
+    plt.scalar_field(dh.gather_array('phi', slice_obj).squeeze())
+    plt.title("φ")
+    plt.colorbar()
+    plt.subplot(1, 3, 2)
+    plt.title("T")
+    plt.scalar_field(dh.gather_array('T', slice_obj).squeeze())
+    plt.colorbar()
+    plt.subplot(1, 3, 3)
+    plt.title("∂φ")
+    plt.scalar_field(dh.gather_array('phidelta', slice_obj).squeeze())
+    plt.colorbar()
+```
+
+%% Cell type:code id: tags:
+
+``` python
+init()
+plot()
+print(dh)
+```
+
+%% Output
+
+        Name|      Inner (min/max)|     WithGl (min/max)
+    ----------------------------------------------------
+           T|            (  0,  0)|            (  0,  0)
+       T_tmp|            (  0,  0)|            (  0,  0)
+         phi|            (  0,  1)|            (  0,  1)
+     phi_tmp|            (  0,  0)|            (  0,  0)
+    phidelta|            (  0,  0)|            (  0,  0)
+    
+
+
+
+%% Cell type:code id: tags:
+
+``` python
+if 'is_test_run' in globals():
+    time_loop(2)
+    assert np.isfinite(dh.max('phi'))
+    assert np.isfinite(dh.max('T'))
+    assert np.isfinite(dh.max('phidelta'))
+else:
+    from time import perf_counter
+    vtk_writer = dh.create_vtk_writer('dentritic_growth_large', ['phi'])
+    last = perf_counter()
+    for i in range(4):
+        time_loop(100)
+        vtk_writer(i)
+        print("Step ", i, perf_counter() - last, dh.max('phi'))
+        last = perf_counter()
+```
+
+%% Output
+
+    Step  0 19.713090835999992 1.0
+    Step  1 19.673075279000045 1.0
+    Step  2 19.696444219 1.0
+    Step  3 19.752472744999977 1.0

tests/test_printing.py

+import pytest
+import re
+import sympy as sp
+
+import pystencils
+from pystencils.backends.cbackend import CBackend
+
+
+class UnsupportedNode(pystencils.astnodes.Node):
+
+    def __init__(self):
+        super().__init__()
+
+
+@pytest.mark.parametrize('type', ('float32', 'float64', 'int64'))
+@pytest.mark.parametrize('negative', (False, 'Negative'))
+@pytest.mark.parametrize('target', (pystencils.Target.CPU, pystencils.Target.GPU))
+def test_print_infinity(type, negative, target):
+
+    x = pystencils.fields(f'x:  {type}[1d]')
+
+    if negative:
+        assignment = pystencils.Assignment(x.center, -sp.oo)
+    else:
+        assignment = pystencils.Assignment(x.center, sp.oo)
+    ast = pystencils.create_kernel(assignment, data_type=type, target=target)
+
+    if target == pystencils.Target.GPU:
+        pytest.importorskip('cupy')
+
+    ast.compile()
+
+    print(ast.compile().code)
+
+
+def test_print_unsupported_node():
+    with pytest.raises(NotImplementedError, match='CBackend does not support node of type UnsupportedNode'):
+        CBackend()(UnsupportedNode())
+
+
+@pytest.mark.parametrize('dtype', ('float32', 'float64'))
+@pytest.mark.parametrize('target', (pystencils.Target.CPU, pystencils.Target.GPU))
+def test_print_subtraction(dtype, target):
+    a, b, c = sp.symbols("a b c")
+
+    x = pystencils.fields(f'x: {dtype}[3d]')
+    y = pystencils.fields(f'y: {dtype}[3d]')
+
+    config = pystencils.CreateKernelConfig(target=target, data_type=dtype)
+    update = pystencils.Assignment(x.center, y.center - a * b ** 8 + b * -1 / 42.0 - 2 * c ** 4)
+    ast = pystencils.create_kernel(update, config=config)
+
+    code = pystencils.get_code_str(ast)
+    assert "-1.0" not in code
+
+
+def test_print_small_integer_pow():
+    printer = pystencils.backends.cbackend.CBackend()
+
+    x = sp.Symbol("x")
+    y = sp.Symbol("y")
+    n = pystencils.TypedSymbol("n", "int")
+    t = pystencils.TypedSymbol("t", "float32")
+    s = pystencils.TypedSymbol("s", "float32")
+
+    equs = [
+        pystencils.astnodes.SympyAssignment(y, 1/x),
+        pystencils.astnodes.SympyAssignment(y, x*x),
+        pystencils.astnodes.SympyAssignment(y, 1/(x*x)),
+        pystencils.astnodes.SympyAssignment(y, x**8),
+        pystencils.astnodes.SympyAssignment(y, x**(-8)),
+        pystencils.astnodes.SympyAssignment(y, x**9),
+        pystencils.astnodes.SympyAssignment(y, x**(-9)),
+        pystencils.astnodes.SympyAssignment(y, x**n),
+        pystencils.astnodes.SympyAssignment(y, sp.Pow(4, 4, evaluate=False)),
+        pystencils.astnodes.SympyAssignment(y, x**0.25),
+        pystencils.astnodes.SympyAssignment(y, x**y),
+        pystencils.astnodes.SympyAssignment(y, pystencils.typing.cast_functions.CastFunc(1/x, "float32")),
+        pystencils.astnodes.SympyAssignment(y, pystencils.typing.cast_functions.CastFunc(x*x, "float32")),
+        pystencils.astnodes.SympyAssignment(y, (t+s)**(-8)),
+        pystencils.astnodes.SympyAssignment(y, (t+s)**(-9)),
+    ]
+    typed = pystencils.typing.transformations.add_types(equs, pystencils.CreateKernelConfig())
+
+    regexes = [
+        r"1\.0\s*/\s*\(?\s*x\s*\)?",
+        r"x\s*\*\s*x",
+        r"1\.0\s*/\s*\(\s*x\s*\*x\s*\)",
+        r"x(\s*\*\s*x){7}",
+        r"1\.0\s*/\s*\(\s*x(\s*\*\s*x){7}\s*\)",
+        r"pow\(\s*x\s*,\s*9(\.0)?\s*\)",
+        r"pow\(\s*x\s*,\s*-9(\.0)?\s*\)",
+        r"pow\(\s*x\s*,\s*\(?\s*\(\s*double\s*\)\s*\(\s*n\s*\)\s*\)?\s*\)",
+        r"\(\s*int[a-zA-Z0-9_]*\s*\)\s*\(+\s*4(\s*\*\s*4){3}\s*\)+",
+        r"pow\(\s*x\s*,\s*0\.25\s*\)",
+        r"pow\(\s*x\s*,\s*y\s*\)",
+        r"\(\s*float\s*\)[ ()]*1\.0\s*/\s*\(?\s*x\s*\)?",
+        r"\(\s*float\s*\)[ ()]*x\s*\*\s*x",
+        r"\(\s*float\s*\)\s*\(\s*1\.0f\s*/\s*\(\s*\(\s*s\s*\+\s*t\s*\)(\s*\*\s*\(\s*s\s*\+\s*t\s*\)){7}\s*\)",
+        r"powf\(\s*s\s*\+\s*t\s*,\s*-9\.0f\s*\)",
+    ]
+
+    for r, e in zip(regexes, typed):
+        assert re.search(r, printer(e))